WorksMathematics
FRCI: An R Package for fuzzy and randomized confidence intervals
Abstract
Fuzzy confidence intervals provide a flexible framework for statistical inference by representing uncertainty through membership functions rather than fixed interval endpoints. Closely related to randomized confidence intervals, they offer an alternative approach for dealing with the conservativeness that frequently arises in discrete statistical models. Despite their attractive theoretical properties, the practical use of fuzzy and randomized confidence intervals has been limited by the lack of readily available software implementations. This paper introduces FRCI (Fuzzy and Randomized Confidence Intervals), an R package available on CRAN for constructing and visualizing fuzzy and randomized confidence intervals. The package implements confidence interval procedures for parameters of the normal, Bernoulli/binomial, and Poisson distributions, including a recently proposed method based on reference value choice, Geyer-Meeden, Agresti-Coull, and score methods. The main function computes the membership function associated with a selected confidence interval procedure, while dedicated graphical tools facilitate visualization and comparison of fuzzy confidence intervals. We describe the statistical background underlying the implemented methods, discuss the package design, and illustrate its use through several examples. The FRCI package provides researchers, practitioners, and students with an accessible and unified environment for the study and application of fuzzy and randomized confidence intervals.